After scaling cluster from 0 to x, external IP address times out about helm-chart HOT 4 CLOSED

this is .... strange.... :-)

Thanks for the careful debugging. We definitely don't want that external IP to change (I feel like it shouldn't since if you scale the cluster down to 0 you should still be paying for the IP address).

@aculich do you have any experience scaling clusters up/down in this way? Ever experienced instability in the network configuration as a result?

from helm-chart.

I don't think the external address ever changed, as that was held for me by google cloud. Certainly the reported service external IP was the same as before.

And when everything restarted, all the k8s components did appear to be pointing to the right ip addresses. It seemed like it was jupyter hub that was source of the problem, although I couldn't see what.

from helm-chart.

hmm, I've had scaling up and down work very well in the past. Let's try to debug this :)

1. Can you provide output of kubectl --namespace=<your-namespace> get pod -o yaml?
2. Can you provide output of kubectl --namespace=<your-namespace> describe pod?

Based on these, we can tell if the pods are running or not, and what stage they're in. We can then go from there.

from helm-chart.

We've merged this repo with https://github.com/jupyterhub/zero-to-jupyterhub-k8s, and I've move this issue over to jupyterhub/zero-to-jupyterhub-k8s#181. Please continue the conversation there, we wanna help you get this to work!

from helm-chart.